US3508037A

US3508037A - Decimal add/subtract circuitry

Info

Publication number: US3508037A
Application number: US612479A
Authority: US
Inventors: David M Collins; Gary J Iverson
Original assignee: Sperry Rand Corp
Current assignee: Sperry Corp
Priority date: 1967-01-30
Filing date: 1967-01-30
Publication date: 1970-04-21
Anticipated expiration: 1987-04-21

Description

April 21, 1970 Filed Jan. 30, 1967 ITEWI QREETIOT ADDRESSABLE MEMORY READ/ WRITE CIRCUITS D. M. COLLINS ETAL ll Sheets-Shee t 1 Fig.

OPERAND SELECTOR BINARY ARITHMETIC SYSTEM w 4 INSTRUCTION REGISTER I AND ADDRESSING FUNCTION CODE TRANSLATION AND OPERAND ADDRESSING 4a L I I r30 DECIMAL I I I TIMING AND I LOGIC NETWORK i SEQUENCE CONTROL 56 AR|THMETIC SECTION I CONTROL SECTION 20 I2 f I k b y 29 24 23 2l 20 1a I7 15 l4 0' BASIC INSTRUCTION WORD FORMAT Fig. 2

f sf I b y I 29----24 2s --Is |7|5 I4 --o| DECIMAL INSTRUCTION WORD FORMAT INVENTORS DAV/0 M. COLL/MS GAR) J. lVERSO/V ATTORNEY April 21, 1970 D. M. COLLINS ETAL DECIMAL ADD/SUBTRAGT C IIICUITRY Filed Jan. 30, 1967 ll Sheets-Sheet Z F (Y) a mu r302 IAQHQQ) XD XI DECIMAL SUBTRACT I 304 I 506 SELECTS STR'P STRIP zas 7H2 AND STORE 3 4. 77-46 308 V 3IO r3I6 052% 5,6Ns ALIKE YES UL NO f SIGNS ALIKE. SELECTS TH I o4= o4 o4= o4 SET SI- FF 320 322 YE DECIMAL COMP. F (A0) (QOI ADDER 324 (AID) (Q@) AD DER /326 FORM BINARY SUM (so BITS) .iia-

PROPAGATE DECIMAL CARRIES CARRY FROM MOST SIGNIFICANT POSITION -EOC SET DCID- FF 332 TEST SI-FF a DCID-FF FOR END CORRECTION /334 ADD DECIMAL CORRECTION FACTORS- A0 Q0 SI-FF SET 338 340 SIGN T a Dem FF SET END CORLRECT I- 346 344 DEC. COMPLEMENT 8 Q6 SIGN (on RESTORE ZONE BITS April 21, 1970 co L s ET AL 3,508,037

DECIMAL ADD/SUBTRAGT C IRCUITRY Filed Jan. 30, 1967 ll Sheets-Sheet 6 IOO QG-REG.

T7-l 206 Tll,Tl5 I66 A0 SELECT 00 SELECT 00 Al A l DQ-DEC. SELECT Dl-DEC. SELECT I A O Q0 AQ E on I I94 I90 I940 35 9/l0 COMP. DEC. CORR. A0 X0 00 W Aw To 00 66 A ril 21, 1970 D. M. COLLINS ET AL 3,503,037

DECIMAL ADD/SUBTRACT CIRCUITRY FIG. 9b FIG 9c FIG. 9 a

A ril 21, 1970 D. M. COLLINS ET AL 3,508,037

DECIMAL ADD/SUBTRACT CIHCUITRY Filed Jan. 30, 1967 ll Sheets-Sheet ll D0 Dl SEL. DEC. COMP. EN.

D0 DI SEL. DEC. IT. EN.

DID DI SEL. DEC. CORR. EN.

FIG. 8

IO's LOWER. EN.

DI DEC. SEL. O2

STAGE 0 CARRY US. Cl. 235-459 19 Claims ABSTRACT OF THE DISCLOSURE This disclosure relates to an arithmetic system for adding or subtracting a pair of decimal operands, where the decimal digits of each of the operands are expressed in binary coded form. The operands, in addition to the binary signals indicative of the decimal digits, include signindicating signals and format-defining signals. The system operates under programmed direction to generate either the sum or the difference while forming the appropriate sign of the result and while establishing the required format-defining signals. The binary coded decimal arithmetic system incorporates a binary adder and a decimal network in a binary data processing system. A programmable system of adding or subtracting coded operands which express values that cannot be stored in a single set of storage registers is also described.

INDEX BACKGROUND OF THE INVENTION Field of the Invention Description of the Prior Art SUMMARY OF THE INVENTIO DESCRIPTION OF THE D RAWINGS.

CONVENTIONS SYSTEM OPE RATION The Data Processing System Instruction Formats Number System Decimal Operands" Decimal Repertoire of Instructions End-OlT-Carry and Overflow Designators Decimal Arithmetic (a) Decimal addition (b) Decimal subtraction Multiple-precision Decimal Arithmetic Sign of Zero Special Case DESCRIPTION OF PREFERRED EMB ODIMENT System Consideration Decimal Network CLAIMS BACKGROUND OF THE INVENTION Field of the invention This invention relates to an improved data processing system, and more particularly, it relates to digital computers and similar types of machines which are useful for performing computation, data processing, and other related functions. Specifically, the invention described is that of an arithmetic system for adding or subtracting coded operands by utilizing a binary adder to forman intermediate partial result, which in turn is operated on by modification circuitry for generating an ultimate coded result. In the preferred embodiment, the digits of the operand are represented in binary coded decimal.

Description of the prior art Many digital computers are designed to handle data words, otherwise referred to as operands, of substantial length, thereby providing the desired data handling ca- United States Patent O 3,508,037 Patented Apr. 21, 1970 pability and providing adequate flexibility of operation in response to the instruction commands commensurate with the design and the application of the machine. It is frequent, however, in such machines that data items to be handled only require a fraction of the total of available capacity of each memory register as defined by the word length. Therefore, it would be unduly wasteful of the available memory space to store such short items in individual addressable registers. Typical of such items are constants, alpha-numeric symbol representations, and other coded values or the like. In the preferred embodiment to be described below, the data work is defined by the capacity of a single addressable memory register and is comprised of thirty-bits. In the system in which the subject invention is to perform, a useful word segment is six-bits. These six-bit portions are referred to in the prior art and herein as bytes. Coincidentally, a desirable mode of expressing alpha-numeric characters in a coded form requires six-bits of storage space. The latter format is referred hereinafter as Fieldata.

Due to the particular adaptability of the binary number system to switching logic, many prior art machines are constructed in a manner to perform their arithmetic operations in the binary number system. These machines are often referred to as binary digital computers. It was recognized at an early stage, that many persons, due to their training, are used to thinking in the decimal number system and the binary number system creates certain problems of adaptability for these persons not having a particular schooling in that regard. Accordingly, it was established at an early time that the user could consider the operations to be in decimal whereas the internal operation of the computer could be made in terms of binary arithmetic. To accommodate this dual radix operation, the binary coded decimal format was developed. In the binary coded decimal format, four binary digits are utilized to represent a single decimal digit. Many systems have been developed based on this binary coded decimal manipulation system.

One of the ultimate goals in all machine constructions is to maximize the utilization of memory cells and registers since this normally is an expensive portion of any data processing system. It is of course apparent as it was in the prior art, that four binary digits can express sixteen possible binary digit combinations and is excessive for the requirements of uniquely expressing ten decimal digits. Accordingly, the prior art systems recognize various ways of accommodating arithmetic Operations wherein binary coded decimal digits were used. Due to the fact that the binary coded decimals were usually in adjacent positional locations, and due to the necessity of providing correction signals for the unused signal combinations in excess of those required to uniquely express the ten decimal digits, the prior art systems usually operated in a decimal digit serial mode in providing addition or subtraction.

As general purpose data processing systems matured in the business field, it became ever increasingly important to provide a more broad spectrum of total system capability. Accordingly, it became desirable to provide a data processing system wherein binary operations could be performed, decimal operations could be performed, and alpha-numeric symbols could be internally handled in an expeditious manner. To accommodate the handling monly utilized for each character. The binary coded decimal signals require only four binary digits but fit within the byte framework. Of course the binary number system can be built into the system without disrupting either the Fieldata or binary coded decimal modes of operation. It should be understood that these three types of data handling formats are not generally used simultaneously, though they may all be used within a single series of data manipulations.

It is a primary objective, therefore, of this invention to provide an improved data processing system wherein the user has the option to select binary arithmetic, or binary coded decimal arithmetic operations within the framework of the system which permits the handling of alpha numerical coded data in a Fieldata format.

To provide further system versatility, it is desirable in large scale data processing systems to provide a capability of adding or subtracting decimal numbers which are of a numerical value greater than that which can be expressed in a single set of operands. Accordingly, the apparatus of this invention includes control circuitry which can be utilized to manipulate coded decimal operands of a size exceeding that of a pair of memory registers. It is, therefore, another primary objective of this invention to provide an improved arithmetic system wherein binary coded decimal operands having a total numerical value in excess of that which can be expressed in a single set of memory registors can be added or subtracted as desired.

It is of course common in arithmetric operations to be forced to manipulate both positive and negative numbers. It is another primary objective, then, to provide an improved data processing system wherein both binary and binary coded decimal operands can be expressed as either positive or negative values and can be handled for addition or subtraction in the improved arithmetic section, which will be described below.

SUMMARY OF INVENTION This invention comprises an improved data processing system wherein binary and decimal arithmetic can be programmed and executed. Further, the invention provides for arithmetically handling coded operands, wherein other than numerical data signals are included. The prior art suggests systems for performing arithmetic operations on binary operands, or decimal operands, or binary coded operands, or operands wherein a set of variable radix conditions can be handled. In this regard it is desirable to provide a system wherein two or more of these types of operands can be operated on arithmetically. The Arithmetic Section of this invention can accommodate binary and binary coded decimal operands, and utilizes the binary adder alternatively for operating on a pair of binary operands or a pair of binary coded decimal operands which have a plurality of decimal digits represented therein. The coded decimal digits are in predetermined positional locations and include non-numerical signal portions for defining format and sign, where non-numerical is used to differentiate between signals which represent the format and sign, as opposed to numerical for designating the decimal digits. These non-numeric portions are intermediate adjacent coded decimal digits. The binary adder operates on the coded decimal digit portions of the operands in parallel so that an intermediate result is generated at a high computational rate. The system then operates to evaluate the partial resultant to add in factors which account for the unused binary digit combinations that result between those required to express decimal digits uniquely and the 16 possible combinations available from the fourbits provided for expressing each decimal digit. These factors, in combination with the required decimal carries or borrows, result in the correct sum or difference value being formed. Having determined the correct sum or difference, the non-numerical format defining signals are inserted therein. Additionally, this invention provides for performing so-called multiple-precision arithmetic operations on binary coded operands. For purposes of this discussion multiple-precision 'will refer to addition or subtraction of a pair of operands which cannot be stored in a single set of memory registers. This mode of operation requires more than one pass through the Arithmetic Section. Further, it requires circuitry for remembering carry or borrow information between segments that are to be manipulated in subsequent passes.

DESCRIPTION OF DRAWINGS FIGURE 1 is a simplified block diagram which illustrates the operational relationship of the primary elements of'the data processing system which incorporates this invention; FIGURE 2 illustrates the format of the basic instruction; FIGURE 3 illustrates the format of the special condition instruction for decimal instructions; FIGURE 4a through FIGURE 4f illustrate the logic circuit symbolic representation and functions for the embodiment described; F-IGURE 5 illustrates the format of the decimal operands; FIGURE 6 is a logic flow diagram which defines the sequence of steps for a Decimal Add and a Decimal Subtract operation; FIGURE 7 illustrates the relationship of FIGURE 7a and FIGURE 7b, and FIGURE 7c which together are a logic block diagram of the elements of a data processing system which embodies this invention; FIGURE 8 is a logic block diagram of the control portion of the decimal network; and FIGURE 9 illustrates the relationship of FIGURES 9a through 9d, which together are the logic block diagram of decimal network and includes the circuitry for performing nines complement, tens complement and decimal correction.

CONVENTIONS The following circuitry discussion of the improved arithmetic system will be in terms of block and logic diagrams since the detail operation of the various elements are. well-known. Various registers will be referred to. These registers can be considered to be comprised of a plurality of bistable flip-flop circuits of a type readily available in the commercial marketplace. It will be understood that each flip-fiop is capable of being set to one oftwo possible stable static conditions. It will further be understood that both the true and the complement value of the existing static state are available as output signals. For example, if a voltage level indicative of a true value of a digit is applied and set into the flip-flop, the voltage level indicative of the true value will be available at the Set (S) output terminal, and a voltage indicative of the Complement value will be available at the Clear (C) output terminal. Alternatively, if the flip-flop is cleared, a voltage indicative of the cleared state will be available at the C output terminal and the complement voltage value will be available at the T output terminal. Each register for this embodiment is comprised of 30' stages designated as 2 for the least significant to 2 for the most significant digit positions. Stage 2. is the binary operand sign indicating stage and 2 is decimal operand sign indicating stage. A logical 1 signal represents a negative binary operand and a logical 0 signal represents a positive binary operand. The reverse sign values are used for decimal operands. Note the distinction between the sign of a binary operand and a binary coded decimal operand.

This embodiment makes use of logical connective circuits of types well-known in the art. The logical AND circuit is denoted by a block with the letter A at the center thereof and represents a logical connective wherein both input terminals must receive like active signals in order for a similar active signal to be available at its output terminal. The logical OR circuit is designated with OR in the center of a block and operates to provide an active signal at its output terminal when any or all of its input terminals receive an active signal. An Inverter circuit is illustrated as a block with an N at the center thereof and operates in a manner well-known to invert the level of the signal applied to its input terminal. Other logic elements are described and defined in FIGURE '4a through FIGURE 4 along with the truth-tables for defining the respective logical operations. These circuits are wellknown and are commercially available, or can be constructed from commercially available components. The circuit operation is obvious to one skilled in the art, and accordingly, a detailed description of the respective operations will not be made, since this would not appear to add to an understanding of the invention. Each of the circuits illustrated performs the logical connective function shown within the symbol. It is understood that each of the elements illustrated in FIGURES 4a through 4 include an inverting signal amplifying circuit at the output. This is reflected in the truthtables. FIGURE 4a is a high-fanoutdriver (HFD) OR-Inverter circuit wherein the output circuit is an inverting amplifier capable of driving up to 60 logic circuit input terminals. When only a single input is used, it becomes an inverter amplifier. In the truthtables a illustrates a dont-care condition and can be any combination of ls and Us. The legend defines the half-arrow and bare-line line terminations utilized in describing the logic arrangements. It is, of course, understood that other types of logic configurations could be utilized in implementing the present invention; however, those shown herein have been found to be advantageous both with regard to cost and operational rates.

In addition to the line connections used in the logic diagrams, the line connections between the various elements are of two basic types. A set of parallel lines terminated by an arrowhead indicates parallel conductive paths for carrying a plurality of data signals. In block diagrams, a single line terminated by a full arrowhead normally indicates a conductive path for carrying control signals. It should be recognized of course that there is no pai'ticular distinction between what is termed a control signal or a data signal other than it is convenient for purposes of discussion to distinguish therebetween. Further, data signals can be utilized for performing control functions. The interconnection lines are terminated by full arrowheads, and they denote the direction of signal flow and can be considered to be points of circuit interconnection.

For purposes of this discussion, the terms logical 1 and logical O or simply 1 or will be utilized to designate the appropriate signal level which is indicative of the binary 1 or 0 value. For example, when considering a two input AND, and ls are considered to be active and available on both inputs, a 1 will be available at its output signal. This numerical designation of these signals is felt to be advantageous in that the precise signal levels will vary depending upon the circuitry chosen for the implementation.

In referring to an address in the addressable memory, the letter Y is utilized. To represent the operand stored at address Y, the designation is (Y), with the parenthesis indicating the operand. The designation (Y) is utilized to represent the 1s complement of the operand stored at address Y. Binary coded decimal operands, and the 9s and lOs complement thereof will be described below.

SYSTEM OPERATION The Data Processing System ,as add, subtract, multiply, store, divide, and the like. In

a stored program calculator the list of instructions (program) which are to be performed are normally stored in the Memory Section shown enclosed in dashed block 10, from whence they are individually read out for execution.

The order of reading out the instructions can be controlled by an Instruction Address Counter (not shown) for selecting sequential instructions, or can be of the type where each instruction designates the address of the next instruction to be executed, or a mixture of each. The mode of control is mentioned generally and will not be described in greater detail since it does not form a part of the invention. In the course of operation, the instructions are normally read into the Control Section, shown enclosed in dashed block 12, and stored in an Instruction Register 14. The instructions normally are comprised of an operation code, that is, the part which defines the function to be performed by the calculator. The instructions also include an address portion for addressing the Memory Section or selecting operands either to be read out therefrom or stored therein. Additional designators are commonly employed to provide address modification by way of indexing or so called B-boxing, for providing for indirect addressing, and for determining whether or not a full length Memory Register or some portion thereof is to be utilized. The instruction word format and the various functions will be described in more detail below. When the instruction resides in the Instruction Register, the operation porton is normally translated by a Function Code Translator and Operand Addressing circuitry 18 for selecting a particular set of control sequence circuits which will provide control signals for guiding the execution of the particular operation to be performed and for addressing the Memory Section. Data processing machines can be of the single-address type, that is, wherein only one memory register of the Memory Section can be addressed and be given instructions; double-address machines wherein each instruction can designate two addresses in the main memory; or three-address machines. The binary coded decimal arithmetic operations which will be described and comprise this invention, will be described within the framework of a single-address machine. It is common in the execution of a program for a resultant operand to be stored both in the Memory Section 10 of the computer and to be retained in the Arithmetic Section, shown enclosed in dashed block 20, for manipulation by subsequent instructions. This is especially common in single-address machines wherein only one operand can be called for in a given instruction. For instance, in a binary coded decimal addition operation, where two operands must be specified, that is, the addend and the augend, it is necessary in a singleaddress machine to first have one of the operands residing in the Arithmetic Section. Subsequently, when the addition operation is called for, the address portion of the instruction designates the second operand which is to be utilized in the binary coded decimal addition operation.

The elements illustrated in FIGURE 1 are considered only generally at this time and will be given a more detailed treatment below. In FIGURE 1, the Memory Section 10 is comprised of storage registers in which the stored programs and recorded data reside, referred to collectively as Addressable Memory 22. The Memory Section can 'be comprised of magnetic core storage registers, plated wire memory registers, bistable flip-flop registers, thin film registers, magnetic drum or disk records or any of the other well-known types of Addressable Memory apparatus. The Memory Section 10 also includes the Read/ Write circuits 24, including amplifiers and address selection circuitry required for recording in a designated storage register and reading out of a designated storage register along

paths

26 and 28 respectively, but for purposes of this discussion are not shown in detail. A detailed operation of the Memory Section 10 need not be considered further, it being understood that Memory Sections are available in the prior art, and the detailed operation of which do not form a part of this invention.

The Control Section 12 is utilized to provide control, for therest of the data processing system. The Control Section interprets and directs the execution of the computer instructions and auxiliary operations. Each computer instruction, except input-output instructions, performs one or more basic operations such as referred to above. Each instruction can also specify certain special auxiliary operations that may modify or extend the basic operation. The Control Section 12 includes Timing and Sequence Control circuitry 30 for timing the execution of each instruction and for accessing the Memory Section. Commonly, an oscillator is used as a primary source of regularly occurring signals. These signals are used to drive timing chains, that is, groups of circuits which provide control pulses at predetermined times and guide the respective data processing operations. The Control Section 12 operates to sequentially read the instruction from the Memory Section into an Instruction Register 14. The Instruction Register stores the instruction being executed at any given time.

The Arithmetic Section functions to perform the arithmetic and data manipulating operations. The data input to the Arithmetic Section 20 is along cable 32 into an Operand Selector 34. The form of the operand is controlled in part by the Function Code Translation and Operand Addressing circuitry 18 by way of control line 36 and in part by the Timing and Sequence Control by way of control line 38. Having selected the form of the operand, which will be described in more detail below, the operand will be transmitted along cable 40 to the Binary Arithmetic System 42. Control line 44 determines the nature of the arithmetic operation to be performed and control line 46 supplies the sequential timing for executing the operation. When a decimal operation is requested, a Decimal Logic Network 48 is in communication with the Binary Arithmetic System through interconnecting lines illustrated graphically as

cables

50 and 52. The control of the Decimal Logic Network 48 is accomplished according to the specific decimal operation to be performed by control line 54, and is timed by signals received on control line 56. Decimal addition and decimal subtraction utilizes the Binary Arithmetic System 42 in conjunction with the Decimal Logic Network 48. It is the interrelationship of these elements in the Arithmetic Section 20 which comprises the essential elements of this invention. A functional and more detailed description of the interrelationship of these elements will be set forth below.

Instruction Word Formats Each instruction performed by the computer is designated by a 30-bit instruction word for the embodiment described; and is performed in two sequential steps, namely, interpretation and execution. .When the instruction word is interpreted, the instruction word is analyzed to activate logic circuits, which will perform a specific operation, as mentioned above in a consideration of the Control Section. When the instruction is executed, the actual function designated is performed. The following listing describes the principal events which occur when an instruction is interpreted and then executed:

(a) Interpretation is performed in four basic steps:

Acquire the instruction word from Memory; Determine the operand source;

Determine the operation to performed; and Condition the control logic circuits.

(b) Execution is performed in three basic steps:

Obtain the operand from Memory when required; Perform Arithmetic or logic operation; and Store the results in Memory when required.

FIGURE 2 illustrates the basic instruction word format and comprises five basic parts which are designated as:

fthe function code designator consists of six-bits which, when decoded, specify the major operation to be performed; j-the branch conditiondesignator consisting of three-bits which, when decoded, specify skip or jump conditions, registers, or repeat modifications;

k-operand interpretation designator consisting of threebits which, when decoded, defines the source and form of the operand. The interpretation of k is different for the read, store and replace categories of instructions;

b-index designator consisting of three-bits which, when decoded, specifies the selected B-Register (Indexing Registers) to be used in conjunction with the modification of the address portion y; and

y-operand, or operand address designator, which consists of lS-bits.

Depending on k, y may represent an operand as it exists,

or an operand address. When y designates the address of an operand, the contents of the designated B Register as determined by the b designator is added to the y portion of the instruction to form an absolute address Y.

FIGURE 3 illustrates the special condition instruction word format for decimal instructions and comprises four basic parts:

fthe function code designator is set to a special code of 77 for decimal instructions, and causes the designator portions to be interpreted in a special manner;

sf-the subfunction designator consists of 6-bits, 18 through 23, and are interpreted as subfunctions of the basic decimal operation. The various subfunctions defines the operations and will be listed and described below.

bindex designator which specifies the selected B-Register as described above; and

yoperand, or operand address designator consists of 15- bits which, when combined with the selected contents of the B-Register, specifies address Y which is the most significant half of the selected decimal operand. A special instruction format is provided for input/output instructions, but will not be discussed since it does not provide a part of the subject invention.

Number System In the binary coded decimal (BCD) number format, four binary digits (bits) are utilized to express each decimal digit. Table I illustrates the binary equivalent of the decimal digits 0 through 9.

TABLE I 0:0000 1:0001, 2:0010 3:0011

In the embodiment of the subject invention, logic circuitry is not incorporated for detecting illegal characters that may be presented to the Arithmetic Section. The illegal characters would be the remainder of the sixteen signal combinations that can be expressed in four binary digits. These illegal codes would be 1010 through 1111 For this embodiment, the presentation to the Arithmetic Section of binary coded signal combinations designated as being illegal may result in an incorrect arithmetic result.

During the execution of decimal operation, two types of complement values are required. These types of complem'ent values are the 10s complement and the 9s complement. For purposes of the discussion of this invention, the 10s complement will be defined as the numerical value generated by taking the 9s complement of each digit in the binary coded decimal operand with the exception of the least significant digit, and the 10s complement of the least significant digit. The 9s complement of an operand is defined as the numerical value generated by taking the 9s complement of each digit including the least significant digit of the original operand. Table 11 illustrates the corre spondence between an originally presented operand and each of the just defined complementary forms.

TABLE II D 9s 10s D 9s 10's 9 *10 or 4 8 9 3 4 7 8 2 3 6 7 1 2 5 6 O 1 *For internal operations, the s complement of zero must be generated as 10 (10102) however, in the final result the 10s complement is generated as zero. This generation or the lOs complement of zero causes a special condition which will be described in more detail below.

By way of example, the following binary coded decimal operand is illustrated with each of the types of complementary designations described above:

EXAMPLE 8 4 3 7 1 9 0 5 2 6 (Original Operand) 1 5 6 2 8 O 9 4 7 3 (9s Complement) 6 2 8 0 9 4 7 4(10s Complement) Utilization of each of the foregoing complementary forms will be discussed in conjunction with the implementation of this invention and their relevance defined more fully below. It will of course be apparent that for the example illustrated above, the binary representation can be derived by substituting the four binary coded digit representations illustrated in Table I into the digit positions for the example.

Arithmetically, the 9s complement of a given digit can be calculated by subtracting the digit from 9. Similarly, the 10s complement of a digit can be formed by substracting a digit representation from 10. As an alternative to the subtraction process, the bit configuration of the binary coded decimal digit arrangement can be evaluated and a system constructed wherein the binary coded decimal representation for the 9s or 10s complement can be generated directly without requiring the execution of the subtract operation. It is the latter system that is utilized in the embodiment of this invention and will be discussed in more detail below.

Decimal Operands All decimal operations within the embodiment of this invention are performed on signed IO-digit operands contained in two 30-bit computer words. Decimal operands having a range of values from 9,999,999,999 through +9,999,999,999 are permissible. FIGURE 5 illustrates the format of the pair of 30-bit computer words which represent a binary coded decimal operand. Operands that are stored in the Arithmetic Section are stored with the most significant half of the operand in the 30-bit A Register and the least significant half of the operand stored in the Q Register. Decimal operands that are stored in the Memory Section 10 are stored in the manner such that the most significant half is stored at Memory Address Y and the least significant half is stored at Memory Address Y+1. The binary coded decimal operands are arranged in positional notation, with the position in the Register being indicative of the value of the respective digits.

Digit positions

0, 1, 2, 3 and 4 are retained in the least significant half Storage Register (Y+1) and digits 5, 6, 7, 8 and 9 are retained in the most significant half Storage Register (Y).

The format defining bit positions are referred to as Zone bits and are designated as Z. The Zone bits are intermediate the binary coded decimal digits, and are a function of the type of coding system designated in the operand. It will be recalled from above, that this system utilizes Fieldata, which is designated by a Z value of 11 and a straight binary coded decimal coding system wherein the Z value is indicated as 00 In performing the arithmetic operations, the Zone bits of the operand stored in the Memory at addresses Y and Y+1 are removed and ignored during the arithmetic operation. The Zone bits of the operand stored in the Arg) Register and the Q Register are removed and saved prior to the initiation of the arithmetic operation. The Zone bits to be utilized in the result are the same as the originally saved Zone bits.

Each decimal operand provides a magnitude of a 10- digit number plus a sign indicating signal. A single bit position is utilized for designating the sign, and for the embodment illustrated, bit position 2 in the least significant half is utilized. The sign bit (S) is interpreted such that a signal indicative of a 1 is utilized to indicate a positive operand and a signal indicative of a 0 is utilized to define a negative operand. It will be noted that this is just the reverse of the sign indication for the usual binary arithmetic. During the decimal arithmetic operation the sign of the operand in A- Q is extracted and stored, as is the sign of the operane read from the Memory Section registers Y and Y+1. The values of the respective signs of the operands is utilized in determining the ultimate sign of the result. This will be described in more detail below.

Instruction Repertoire The following decimal instructions are included in the repertoire of the illustrative embodiment of the invention:

77-l0 Decimal Test 11 Decimal Add -12 Decimal Subtract -l3 Decimal Compare Greater -14 Decimal Complement -l5 Decimal Add With Carry -16 Decimal Subtract With Borrow -17 Decimal Compare Less Y Skip Condition 2 Skip if Overflow Designator is Set. 2 Skip if Overflow Designator is Not Set.

2 Skip if (A) (Q) (Decimal) is Positive (Q4=l). 2% Sk p f (A s) (Q,) (Decimal) =i0 (Sign and Zone bits are ignored). 2L" Sk p f (A) (Qo) (Decimal) 2 Sk p if D (A3': ):|:0.

ggup if 7 ias-i910.

is Negative (Q04=0)- Skip if Da q 2l lB):|:O- Skip f D (Ad 2124)-l:0- 2 Skip il (Ao) (Q) (Decimal) i0 (Sign and Zone bits are ignored).

When more than one skip condition is specified, the skip is made when any one or more of the specified skip conditions are satisfied.

77-11 DECIMAL ADD: The decimal contents of Y, Y+1 are added to the decimal contents of A Q. The result is stored in A Q 5 along with the Zone bits of the original A Q operand. During the execution of this instruction, the states of the End-Off-Carry and/or the OVERFLOW designators may be changed. When it is necessary to complement one operand (A) (Qqb) or the result, the tens complement is generated. A longer execution time is required when it is necessary to complement the result.

7712 DECIMAL SUBTRACT: The decimal contents of Y, Y+1 are subtracted from the decimal contents of A 5 Q 5. The decimal subtract operation is identical to the decimal add operation except for the conditions that determine when it is necessary to complement one of the original operands (A) (Qq5).

77-l3 DECIMAL COMPARE EQUAL: Thle decimal contents of Y, Y+1 are compared with the decimal contents of (A 5) (Qgb) and when they are equal, the next sequential program step is skipped. The contents of A45, Qq5, Y, and Y+1 remain unchanged.

77-14 DECIMAL COMPLEMENT: The decimal complement of the contents of A Q is generated and loaded into A 5 Qqb with the Zone bits of A45 Q unchanged. The least-significant bit position of the instruction (2 of the y field) plus the contents of the specified index register determine whether the nines or tens complement is generated. When this bit is a one, the nines complement is generated and when it is a zero, the tens complement is generated.

77-15 DECIMAL ADD WITH CARRY: The decimal contents of Y, Y+1 are added to the decimal contents of A15 Q 5. The decimal add with carry operation is identical to the decimal add operation except the End-Off-Carry designator is introduced into the correction factor as a carry into the least-significant digit position and when it is necessary to complement the result and/or one of the operands (A (Q), the nines complement is generated. Note that the state of the End-Off-Carry designator cannot be changed until after the correction factor is generated.

77-16 DECIMAL SUBTRACT WI'IH BORROW: The decimal contents of Y, Y+1 are subtracted from the decimal contents of A Qq5. The decimal subtract with borrow is identical to the decimal add with carry except for the conditions that determine when it is necessary to complement one of the original operands (A) (Qqb).

77-17 DECIMAL COMPARE LESS: The decimal contents of Y, Y+1 are compared with the decimal contents of Aq5 Q and when the Y, Y+1 operand is greater than the A Q operand, the next sequential program step is skipped. The contents of A, Qq5, Y, and Y+1 remain unchanged.

End-Oif-Carry And Overflow Designators The Internal Function Register (IFR) is utilized to provide an orderly return to a program which has been momentarily disrupted to honor an interrupt, and to store special designator conditions. Those designators which are relevant to the handling of decimal operands will be described in detail in conjunction with FIGURE 7. As illustrated, two designators are utilized during the execution of decimal instructions. Bit position 2 of IFR is designated f4 or DC F/F, and is used to indicate the state of a decimal carry referred to as End-Oif-Carry (EDC). The conditions under which the End-Off-Carry is set will be described in detail in the consideration of the operation of the decimal add and the decimal subtract. Bit position 2 of IFR is designated the f5 designator, or DEC OVFL F/F, and is used to indicate an external decimal overflow. Each time a decimal add or decimal subtract, with or without carry or borrow, is performed, the End-Off- Carry and overflow conditions within the Arithmetic Section are checked and the corresponding designator in IFR is set or cleared. The only times when the states of these two designators can be altered is when decimal add or subtract instructions are executed, or when the IFR is initially loaded. Other functions of IFR are not relevant to this invention and will not be described.

Decimal Arithmetic Decimal arithmetic is performed in the described embodiment by utilizing the available, 60-bit, 1s complement, subtractive binary adder and a special decimal logic network which generates decimal complements and correction factor constants used to correct the binary sums generated. The two basic decimal arithmetic operations (addition and subtraction) are described in the following paragraphs.

Decimal Addition Decimal addition is performed on two ten-digit operands, one in A, Q15 and the other (Y, Y+1) in memory. The Zone bits and Sign of A, Qq5 are stored in auxiliary registers and the Zone bits of Y, Y+1 are stripped as the operand is transferred into the Arithmetic Section. Since the operands contain the magnitude of a ten-digit number plus a Sign bit, the sequence of operations varies slightly depending on whether the two operands have like or unlike signs.

Operands having like signs may be added without modifications. After the Zone and Sign bits of (A 5) (Qq5) have been stored, they are stripped from the original operand, which is added to (Y), (Y+1) (less Zone and Sign bits). This is a binary add operation. Since the Zone bits have been stripped, and added as zeros, the addition is performed on ten pairs of four-bit numbers with no possibility of a carry propagating from one character to another. The term carry is used here only as an analogy. Since the main adder is a subtractive adder, borrows are propagated between characters; however, the adder output is the same as would be generated by an additive adder.

Since the original digits range only from zero through nine (11 the binary coded sum digits, also referred to as partial resultant, may range from zero through 18 (22 Each character in the partial resultant is used to generate correction factor constants which, when added to the partial resultant, will generate the correct sum in the proper coded decimal format. The correction factor generator also handles all decimal carries between digits. Correction factors are required when a sum digit exceeds the maximum allowable decimal digit (nine). When this condition is detected, a carry to the next digit is generated and a constant is generated which, when added to the original binary coded sum-digit, causes the final sum to skip over the intermediate illegal values (12 through 17 Carries and corrector factor constants are generated as illustrated in Table III.

TABLE III The Following Correction Factor is Generated if, From The Preceding Digit, There is When the First Sum-Digit is No Carry A Carry Less than 9.. 0 1 Equal to 9.-.. 0 *7 Greater than 6 *7 f carry is generated to the following digit position when the sum digit exceeds nine and when the sum digit equals nine and a calry is generated by the preceding digit.

tion is discarded and the resulting sum is recombined with the original Zone bits of A Q.

Example I illustrates Decimal Addition with like signs. The original operands are in proper Fieldata format and are shown in octal notation. The binary system is also shown and the result is shown in decimal as well as Fieldata format.

EXAMPLE I.DECIMAL fiAgIg/ITION WITH LIKE SIGNS A. In Octal:

(1a) (Y) (Y+1)=6167706662 7164636701 (2a) (AqS) (Q)=6660617064 7167677000 Strip and store Zone bits and Sign of (A) (Q76) (3a) 60606060606060606000 The (Y) ,d(Y+1) and (Alt) (Q 3) operands with Zone bits and Sign strippe (40.) 0107100602 1104030701 Stripped (Y) (Y+1) (5a) +0600011004 1107071000 Stripped 11 (Q) (6a) 0707111606 2213121701 artial Sum The correction factor constants are generated and added to the partial sum. (7a) 0707111606 2213121701 artial Sum (8a) 0001070601 0707070600 Correction (98.) 0710202407 3122212501 Second Su Combine Zone bits and Sign stripped from (11) (Q,-) with Sum. (10a) 0710000407 1102010501 St ipped Second Sum (11a) 6060606060 6060606000 Zone and Sign (123.) 6770606467 7162616501 Result B. xpressing the foregoing in Bina y:

D3 D2 D1 D di) 11/0110/11/0000/11/0001/11/1000/11/0100 (2b) 8 D7 6 D 4 3 2 D1 D0 Stripped Zone bits and Sign of (A 5) (Q) (3b) 110000110000110000110000110000 110000110000110000110000000000 Operands wit Zone bits and Sign St ipped.

(Y) =00/01001/00/0711l/00/ l8000/ 0O/06110/00/ 0010 {(Y+1) =00/1001/00/0100/00/0011/00/0111 00/0001 l(A4 00/2110/00/3000/00/0001/00/1000/00/0100 Addition of two decimal numbers having unlike signs may be accomplished by adding the number with the greatest magnitude to the tens complement of the other. To save the time required for a magnitude compare operation, the System arbitrarily complements the operand in 14 AQ whenever unlike signs are detected. The addition operation then proceeds in the same manner as addition with like signs.

When the operand having the smallest magnitude is complemented, the number generated will exceed the uncomplemented operand by an amount sufficient to cause an End-OfiCarry to be generated. When the operand having the greatest magnitude is complemented, no End- Off-Carry is generated and the resulting decimal sum (second sum) is the decimal complement of the desired result. In this case, the tens complement of the sum is generated and the result is assigned the Sign of the AQ operand. When an End-Off-Carry is generated, the sum is correct and is assigned the opposite Sign as that of the original AO operand. Note that External overflow cannot occur (and will never be indicated) when operands having unlike signs are added; however, the End-Off- Carry is retained to facilitate the programmed, multipleprecision decimal arithmetic, that is, operations performed on operand fields greater than ten digits in length, as is described in succeeding paragraphs.

Example II illustrates Decimal Addition with operands having unlike signs. Operand magnitudes are the same as those used in the preceding example, only the sign of (A45) (Qo) has been changed to positive. Note that if the roles of the two operands used in the Example II were reversed, the operand with the smallest magnitude would be complemented, an End-Oif-Carry generated, and the second sum would be the desired result. In this case, the generated End-Otf-Carry does not indicate External Overflow, but is retained for future use under program control. Only the Octal representation is illustrated since it is felt that Example I clearly sets forth the relationship between zone bits. Sign bit and the Binary representation.

EXAMPLE II.DECIMAL ADDITION WITH UNLIKE SIGNS (6a) Partial Sum: 0420200707 1106051013 The correction factor constants are generated from, and added to the Partial sum.

(70) Partial Stun: 0420200707 1106051013 (8a) Cor. Fact.: 0107060000 0000000106 (9a) Second Sum: 0527260707 1106051121 (No E00 The overflow into the Zone and Sign bit positions is stripped, the 10s complement is generated (since there was no End-Off-Carry), the original Zone bits of A05 Q are restored to their original positions, and, since the operand with the greatest magnitude was comglgmgted, the result is assigned the original sign of (10a) Stripped Sum: (11a) Stripped Sum: (12a) Zone and Sign Bits: 6060606060 6060606060 (13a) Result: (Octal) 6462636262 6063646071 (13b) (Binary 11/0100/11/0010711/0011/11/0010/11/0010/11/0000/11/0011/11/0100/11/0000/11/1001 (13c) (Decimal) Decimal subtraction Decimal Subtraction is identical to Decimal Addition except the conditions for determining that one of the operands must be complemented (before adding) are reversed. When the operands have unlike signs, they are EXAMPLE III.DECIMAISIIETI JIISBTRACTION WITH UNLIKE (Adi) (QqS) 6167706662 7164636701 (Y), (Y+1) 6660617064 7167677060 (Art) (01) Stripped 0107100602 1104030701 (Y) (Y-H) Stripped 0600011004 1107071000 First Sum- 0707111606 2213121701 C01. Fact 000107060 0707070600 Second Sum. 0710202407 3122212501 Stripped Sum 0710000407 1102010501 Zone and Sign Bits. 6060606060 6060606000 Result 6770606467 7162616501 EXAMPLE IV.-DECIMAL SUBTRAOTION WITH LIKE SI GNS (A 5) (Qqfi) 6167706662 7164636761 (Y) (Y+1) 6660717064 7167677060 (11) (61) D (Stripped): 1002010307 0005060211 0600111004 1107071000 First Sum 1602121313 1114151211 Dec. Cor 0601070707 07 07070600 (E-O-C) Second Sum 2403212222 2023242011 Second Sum Stripped 0403010202 0003040011 Zone and Sign Bit 6060606060 6060606040 Result 6463616262 6063646051 No'rE. In Example IV, the sign of the result is different than the sign of the original operands because (Y) (Y+1) (A)(Q); however, Zone bit 2 remains unchanged.

Multi-precision Decimal Arithmetic The Decimal Add With Carry and Decimal Subtract With Borrow instructions are provided to simplify the programming of decimal operations requiring operands greater than ten-digits in length. These two instructions difier from the Decimal Add and Decimal Subtract instructions in the following two ways:

(a) Whenever it is necessary to complement an operand (or result) the 9s complement is generated.

(b) The End-Ofli-Carry from the previous Decimal Add or Subtract (normal or with carry/ borrow) is introduced into the decimal correction factor as a carry into the least-significant digit position.

Note here that the End-Olf-Carry condition is not always the same as the External Overflow condition. When Decimal Addition is performed with like signs or subtraction with unlike signs, the End-Off-Carry does indicate External Overflow. When Decimal Addition is performed with unlike signs or subtraction with like signs, External Overflow cannot occur and the End-Off-Carry indicates the second sum is the desired result. In either case, the End-Ott-Carry is used in the same manner by the Decimal Add With Carry and the Decimal Subtract With Borrow instructions.

When performing a Decimal Add (or Subtract) using operands greater than ten digits in length, the following sequence must be followed:

(a) Divide each operand into ten-digit segments (60- bits per segment) starting from the least-significant digit in the operand. Each segment of each operand shall contain ten digits (unused digits in the most-significant seg- 16 ment must contain zeros) and the sign bit (2 shall be the same in each segment. Each segment to be used as the Y, Y+1 operand must be stored in two consecutive memory locations.

(b) Perform a Decimal Add (or Subtract) Without Carry (Borrow) on the least-significant segments of the two operands.

(c) Store the first sum in memory and perform Decimal Add (Subtract) on the remaining corresponding pairs of operand segments With Carry (Borrow) and store each sum (difference) in memory.

Because there is no Way of determining the relative magnitudes of operands until the final add (or subtract) is completed, the Signs assigned to individual segments of the result may differ after an add With unlike Signs. or

a subtract with like Signs. Each segment add is performed independent of the others, with the exception of the End- Off-Carry propagation and each segment of the result may or may not be complemented and is assigned a Sign depending on the relative magnitudes of each segment pair in the original operands.

The Sign of the result is the same as the Sign of the most-significant segment in the result. Once the Sign is determined, it must be compared with the Sign of each segment in the result. When the Signs are equal, the segment is correct. When the Signs differ, the segment must be complemented by forming the 10s complement for the least-significant segment and the 9s complement for the other segments. Examples V and VI illustrate multiple-precision decimal arithmetic. For simplicity, three digit segments are used.

Example V is of particular interest because it illustrates (l) the significance of the End-Olf-Carry during a. subtract operation and (2) Sign propagation through segments equal to zero.

EXAMPLE V.MULTIPLE-P RECISION DEGIMAL ADDITION Third Add Second Add First Add Segment Two Segment One Segment Zero Add With Add With Normal Carry Carry Add (Adv) (Qdi) 616263 646566 677071 (Y) (Y+1) 607007 666504 636201 Strip-Zone and Strip Zone and Strip Zone and Sign bits and Sign bits and Sign bits and 9s comple- 9s comple- 10's comp1ement (A) ment (14) ment (AqS) (Q)- (Q)- (Q)- 100706 050403 020101 00l007-EOC 060504 030201 First Sums 101715 131107 050302 Cor. Fact 010707 060000 000000 112624 211107 050302 9's complement Second sum is 10 complement second sum 0k (End-Ofisecond sum (no End-Oii- Garry). Strip (no End-01T- Carry). Strip Zone and Sign Carry). Strip Zone and bit positions. Zone and Sign Sign bit posi- Sign opposite bit positions. tions. Sign Sign of (A) Sign equals equals Sign of (Q,). Sign of (A6) 40 (Q) (Q) Final Sums 000305 011107 040610 20113155t and Sign 606060 606040 606060 1 5. Results 606365 617147 646670 TDifiterent Signs Same Signs T The program must now compare with the Sign of the most-significant segment ith the other two segments. The Sign of the first segment is positive so the segment is correct as is; however, the second segment (llmust be complemented (9s) which may be accomplished by means of the Decimal Complement instruction. Final Result" 606365 62 646670 Norm-Checking by long-hand subtraction: +123456789 -087654321 EXAMPLE VI.MULTIPLE-PRECISION DECIMAL SUBTRACTION Third add Second dd First Add Segment Two Segment One Segment Zero Subtract Subtract Normal With Carry With Carry Subtract (Qgb) 616263 646560 646471 (Y) (Y+1) 616263 646561 637067 Strip Zone and Strip Zone and Strip Zone and Sign its and Sign bits and Sign bits and 9's comple- 9s eomple- 10s complement (14) ment (Alp) ment (A05) (Q)- (Q40. 100706 050411 050501 010203 040501 031007 irst sum 111111 111112 101510 C01. Fact 070707 -EOC 070706 010600 202020 202020 112310 (EOC) Second um is 10's comple- Second sum is correct (Endment second correct (End- Oil-Carry). sum (No End- Ott-Carry). Strip Zone Ofi-Carry). Strip Zone and Sign bit Strip Zone and Sign bit positions. and Sign bit positions. Sign opposite positions. Sign opposite Sign of p) Sign equals Si of (Q g of (M) inal Difierences 000000 000000 000002 Zone and Sign Bits 606040 606040 606060 Result 606040 606040 606062 T Same Signs T T Difierent Signs L,

The program must now compare the Sign of the most-significant segment with the other two segments The Sign of the second segment is negative so the segment is correct as is; however, the first segment must e complemented (10's) which may be accomplished by means of the Decimal Complement instruction.

Final Result 60 716350 N o'rE.Ohecking by long-hand subtraction: 123451387 +123450449 by the long-hand method and vice versa. This indicates that, under these conditions, the End-Oif-Carry propagation is actually the propagation of a No-Borrow condition. Note also that this Na-Borrow condition is analogous to a No-Overflow condition while the End-Oif-Carry generated during Decimal Addition with like Signs and subtraction with unlike Signs is analogous to an Overflow condition.

In the Example VI (subtraction) the Sign of the result is correctly indicated by the two insignificant (equal to zero) segments, while the significant (not equal to zero) segment must be complemented by the program to obtain the desired result. Note that the Sign of the most-significant segment is determined by the presence or absence of an End-Off-Carry into that stage and any zero segment receiving an End-Off-Carry will also propagate an End Off-Carry to the next more significant segment. This assures the correct Sign of the most significant segment regardless of the number and location of zero segments in the multi-segment.

Sign of Zero The sign of the decimal result is determined as follows: Opposite of the Sign of (A11) (Q) if:

(a) it is not necessary to complement one of the operands before adding (or subtracting);

(b) it is necessary to complement one of the operands before adding (or subtracting) and no End- OiT-Carry is generated.

Opposite of the Sign of (A) (0 5) if:

(a) it is necessary to complement one of the operands before adding (or subtracting) and an End- Otf-Carry is generated.

In all cases when the result is not zero, the Sign of the result is correct (considering each segment independently during a multiple-precision operation) and can be readily predicted by long-hand computation. When the result is equal to zero, the above stated sign-determining conditions are followed; however, the resulting Sign is not as easily predicted as in the non-zero case. All possible combinations of zero generating operands are tabulated in Table IV along with the Signs of the results. Note that since the At Q operand is always complemented in those cases when it is necessary to complement one operand, the Sign of the result depends not only on the relative Signs of the two operands but depends also on which operand is the A Q operand, the type of operation being performed and whether the operation is with or without carry (or borrow).

TABLE IV.-ZERO RESULT SIGN ASSIGNMENT Addition Without Carry.

Subtraction Without Carry.

Art 0 +13 Y+1-- Subtraction With Borrow. (N o End-Ofi-Oan'y Into Segment.)

+N (N +1)+(N+1) Subtraction With Borrow. +0 0 (End-OfE-Carry Into Segment.)